GPT-6.1 Terra Is Now Live on VEGA

GPT-6.1 Terra is now available on vega.chat, under the “GPT 6.1 Terra” label in the model picker, and through api.vega.chat. The new OpenAI model combines detailed visual reasoning, precise computer use, and more natural writing as a faster, cheaper successor to GPT-5.6 Terra for the tasks that fill a workday.

What OpenAI built GPT-6.1 Terra for

The pitch is practical: a model that reads visual information and turns it into finished work. OpenAI's examples cover drafting proposals, filling out forms, interpreting technical drawings, and editing videos.

For developers, the announcement leans on high-volume automation, like reading incoming invoices and typing the details into accounting software. More specialized cases from the benchmarks: turning architectural drawings into 3D models, producing shot-by-shot time-coded logs for film restoration, and writing code that tracks cell divisions in time-lapse footage.

One caveat on the computer-use numbers: they come from agent-style setups where the model drives desktop apps through tools. In a chat session Terra still works through text and images rather than clicking around for you.

Benchmarks: visual reasoning and computer use

OpenAI's announcement reports gains across several evaluations:

• Agents' Last Exam (professional computer-use tasks): at medium reasoning effort Terra scores 15 percentage points above GPT-5.6 Terra at about one-third the cost, and matches GPT-6 Astra at one-tenth its price per token. At max effort it is over 3× as token-efficient as Claude Opus 5.5 at a quarter of the price, roughly 9× lower cost per task. Tasks also finish over 3× faster than with GPT-6 Astra or GPT-6.1 Sol.


• MMMU-Pro (visual reasoning): beats GPT-5.6 Terra at every reasoning setting and nearly matches GPT-6 Astra and GPT-6.1 Sol at lower cost.


• OSWorld 2.1 (desktop apps): at medium effort, passes GPT-5.6 Terra's best score at a fraction of the cost per task.


• Terminal-Bench 4.0 (command-line coding): 50.3% of tasks solved at max effort, versus 21.5% for GPT-5.6 Terra. OpenAI still recommends GPT-6.1 Sol and GPT-6 Astra for the hardest coding work.


• Terminal-Bench-Science 0.1: at max reasoning Terra beats GPT-6.1 Sol at lower cost and lands within five points of Opus 5.5 at less than a fifth of the estimated cost.

Factuality, safety, and pricing

On OpenAI's internal benchmark of error-inducing prompts, the low-effort error rate falls from 53.6% to 16.7% compared with GPT-5.6 Terra. At maximum effort it drops another 18 points to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol.

OpenAI also reports better scope adherence, more honest disclosure of limitations, and stronger prompt-injection resistance. Terra hits the Critical cybersecurity threshold under the Preparedness Framework and keeps GPT-6 Astra's protections, with details in the system card.

OpenAI's API pricing is $1 per million input tokens, $0.05 cached, $5 output — 50% lower than GPT-5.6 Terra and GPT-6.1 Sol, per the announcement. On OpenAI's own platform the model is live for Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-terra. Those are OpenAI's prices, not VEGA's.

Try GPT-6.1 Terra on VEGA

Pick “GPT 6.1 Terra” in the vega.chat model list and run it on your own documents, drawings, and everyday office tasks. Developers can call the same model through VEGA's OpenRouter-compatible API for agents and automation pipelines.

GPT-6.1 Terra is available in the vega.chat model picker and through api.vega.chat.